Arthritis has a tremendous impact on HRQOL. Thus, measuring HRQOL is an important part of assessing the burden of disease. The National Arthritis Action Plan (NAAP) in conjunction with the Arthritis Foundation supports using the CDC HRQOL modules to increase HRQOL surveillance in people with arthritis [12
]. State and local efforts can then target populations who bear the greatest burden. In addition, the Healthy Days measures are explicitly incorporated into the goals for HRQOL in Healthy People 2010 [13
]. This surveillance helps to target those high-risk individuals who would benefit from medical, self-help, and community-based interventions.
Mili et al. used the CDC HRQOL core module in the general population of 15 states and Puerto Rico and compared individuals with and without arthritis. For the 4-item Healthy Days Core module, participants with arthritis were three times more likely to report their general health as fair to poor and averaged more physical, mental, and overall unhealthy days than participants without arthritis [14
]. Dominick et al. examined the CDC HRQOL modules using Medicare data on 41,467 older adults from Pennsylvania with and without arthritis. They reported the CDC's HRQOL modules were able to distinguish between those with and without arthritis as well as between the different types of arthritis (OA and RA) [15
]. In another study by Currey et al., differences in HRQOL among diagnoses groups (OA, RA, and FM) that were established using condition-specific measures were mirrored on the 4-item Healthy Days Core Module in a clinical population with arthritis [16
]. There appears to growing evidence of the advantages of using the CDC HRQOL modules in the clinic .
A CDC-funded validation study concluded that the CDC HRQOL modules are a reasonable alternative to the SF-36v2, but this study lacked the sample size to evaluate specific disease groups such as arthritis [10
]. A more recent study by Abell et al. used the 2001 BRFSS data from all 50 states, the District of Columbia, and the US territories to assess the relationship between physical activity and HRQOL in people with arthritis. People with arthritis had more unhealthy days compared to those without arthritis. This study's main focus was however not on the psychometric properties of the CDC HRQOL items . Therefore, the results of our study are an important contribution to the conceptualization of HRQOL in people with arthritis.
The underlying factor structure of the CDC HRQOL remained stable across the different patient samples. The samples differed from each other in two distinct ways, i.e., community-based vs. subspecialty and self-reported vs. physician-reported arthritis. The MSK participants are from musculoskeletal subspecialty clinics and they had a physician diagnosis of arthritis. The NC-FP-RN participants are from general practitioners across the state and had self-reported arthritis. Despite these distinctions the CDC HRQOL items remained stable when examined across these populations. Future studies on a population with physician-reported arthritis and physical and mental assessments that are not self-reported would provide further evidence for the validity of the 9-item CDC HRQOL because of potentially more accurate measures.
The unrotated solution suggests that when unconstrained, the mental and physical health-related problems of the CDC HRQOL are so correlated that they are not clearly distinguishable. In this case, a general health factor emerges. This general factor encompasses mental and physical health-related problems and this one-factor solution could be used as a general measure of HRQOL. When rotation is used, the two factors become distinguishable as correlated but distinct factors. A two-factor solution could be used to give more information about the contribution of physical problems and mental health problems to the overall score. These two factors will not be equally correlated for every individual or in every case, so to separate them gives more information.
In a MTMM analysis, the highest correlations are expected between the two different measures of the same trait, while the different trait/same method correlations are expected to be lower. These relationships were confirmed. Although the MTMM correlation was very low for CDC-MH with PCS (-0.35), the CDC-PH and MCS relationship was surprisingly strong (-0.50); this is counterintuitive because correlating different methods and different traits usually results in the lowest correlations. Another counterintuitive finding was the correlation between the two factors on the CDC HRQOL, which was relatively strong (0.58). These relationships suggest that there is a correlation between HRQOL related to physical and mental health problems, and the CDC HRQOL factors seem to reflect this more than the SF-36v2 factors.
Consistent with the factor analysis results of the CDC HRQOL, the general health item seems most closely related to scores on the physical health factor despite the strong correlation between physical and mental health problems. Whereas the inclusion of the general health rating increased the prediction of physical health, it made no difference in predicting mental health as measured on the SF-36v2. Although these results (regression) make the CDC HRQOL appear to be a good approximation of the SF-36v2, the physical and mental health factors are more highly correlated and therefore less distinct than the 'equivalent' SF-36v2 factors. In this study, the CDC PH and CDC MH are proxy measures of a proxy for HRQOL and therefore may include some drift away from validity.
One possible explanation for this drift away from validity is the use of the equal-weighting method employed to create the scale. Although, correlations between the weighted factor scores (from the PCA) and the unweighted means of the CDC PH and CDC MH were >0.99 (p < .0001) for both the overall sample and by sample thus supports the case that the weighted and the unweighted methods of scaling provide similar information. An alternative approach for future research might be to use differentially weighted items based on the relative strength of each item in relation to a standard measure of HRQOL. For example, the Healthy Days Core Module physically and mentally unhealthy days are widely used as single-item global measures of HRQOL. Compared to the global Core Module measures, the Optional Module measures are amendable causes of poor mental or physical HRQOL that allow monitoring of modifications in health programs [6
]. Therefore, future studies might compare differentially weighted Optional Module measures with these global measures and/or with the summary index of unhealthy days.
In this study, there did not appear to be any respondent fatigue or ordering effects. Although, substantial differences would have been important (and, thus, our choice to alternate order the results based on alternating order didn't differ substantially from the results disregarding order. The factor structure for the subscales and the alphas for each subscale did not change according to test order.
The psychometric properties of the 9-item CDC HRQOL support its use in both community-based populations with self-reported arthritis and subspecialty-based populations with physician-reported arthritis. The factor structure of the CDC HRQOL remained stable across both these patient populations. The results of this study are helpful in instances where parsimony of items is important. The 4-item CDC-PH could be used when the primary goal of a project is to measure physical HRQOL; the 3-item CDC-MH could be used when the primary goal is to measure mental HRQOL. Both the CDC-PH and CDC-MH subscales demonstrated good internal consistency. The expected correlations in the MTMM supported the construct validity of the CDC-PH and CDC-MH subscales . In summary, the CDC HRQOL appears to be a valid way to monitor the health-related quality of life of people with arthritis.